Application developer and systems builder shipping production software across AI, media, data, and automation.

I build end-to-end systems solo: native TV experiences, cross-device applications, AI pipelines, APIs, data normalization layers, and the infrastructure that keeps them running.

BS Computer Science. I run Flash AI Solutions, where I turn ideas into working products and reusable engineering systems.

• AI systems with judgment. Multi-stage LLM pipelines, model routing, grounding, fail-closed gates, MCP tooling, and agent workflows built to survive real inputs instead of demo inputs.
• Cross-device applications. Web, desktop, iOS, Fire TV, and Tizen clients with shared contracts, continuity, playback state, and platform-specific UX.
• TV-first interfaces. Kotlin + Jetpack Compose, deterministic D-pad navigation, focus restoration, 10-foot layouts, and performance work for constrained hardware.
• Data + API products. Typed schemas, upstream normalization, identity resolution, REST/MCP interfaces, SQL-backed services, and systems that reconcile messy external data into clean internal models.
• Automation + infrastructure. Python/Node services, monitoring, deployment workflows, Linux/VPS operations, and AI-assisted engineering systems designed for repeatability.

• Flash Props MCP — live sports and esports player-prop data exposed to AI agents through Model Context Protocol, including market metadata, projections, evidence, and line movement.
• tv-focus-navigation-kit — reusable solutions for the hard parts of Compose-for-TV focus: bring-into-view, hero scroll races, cross-zone restoration, and focused-vs-selected state.
• cross-device-continuity-playbook — patterns for keeping playback state coherent across multiple clients using shadow authority, divergence checks, and safe cutovers.
• esports-media-metadata-schema — typed JSON Schema + TypeScript models and an adapter-based normalization layer for reconciling differently shaped sports/esports providers into one canonical model.
• term-hunter-lite — a multi-stage LLM pipeline with fail-closed gates, cost-aware model selection, three-state RDAP checks, and an SSRF-guarded fetch layer.
• ai-agent-project-template — a practical working agreement for AI coding agents covering coordination, verification, secrets, scope, and safe repository changes.

Python · TypeScript · JavaScript · Kotlin · C# · SQL
Node.js · React / React Native · Jetpack Compose · Astro · Power Platform
Postgres / SQLite · Docker · Linux · PM2 · Caddy · REST · MCP · LLM systems

Building production-grade AI-assisted software, cross-device media architecture, and data products that turn noisy upstream information into reliable user experiences.

Portfolio: flashaisolutions.org/work
Email: malone.jaylon@gmail.com